The reduction of the human environmental footprint can be achieved through the use of more efficient motors, such as synchronous reluctance motors (SynRMs). Known for their affordability, SynRMs are increasingly employed in sensorless AC drives. This paper presents a critical comparison of sensorless algorithms based on two nonlinear Kalman Filters (Unscented and Extended). The objective is to highlight the theoretical and practical advantages and drawbacks of each method when applied to the speed control of a SynRM, culminating in a definitive decision on the best choice. The magnetic model of the SynRM is much more nonlinear than that of induction motors, which are predominantly addressed in existing literature. This study aims to fill the gap by answering the question: 'Is the extended Kalman filter still the best choice even in the case of nonlinear electric motors?' The answer comes through a large batch of experiments, including speed and load torque tests, zero-speed standstill starts, parameter sensitivity, and evaluation of computational burdens for both Kalman filter algorithms.